FutureOS：多端共享同一AI代理

One AI agent, everywhere you work — terminal, desktop, mobile, and your chat apps. Rust core.

One AI agent, everywhere you work. Terminal, desktop, mobile, and your chat apps — one Rust core, one agent. Configurable approvals and OS sandboxes. Local-first. Open source. Quick Start • Features • Models • Commands • Troubleshooting ⭐ If FutureOS is useful to you, a star helps others find it. Why FutureOS Choose the execution boundary. Configure file-access approvals and OS sandboxing for agent tools. The desktop defaults to Unrestricted (off); select Manual or Sandboxed before working with untrusted content. See the security model and sandbox guide for defaults and limitations. One backend, every surface. A single gRPC agent drives the terminal UI, desktop app, mobile apps, CLI, and IM bots — same sessions, same memory, same skills, wherever you happen to be. Long runs need engineering, not prompting. The built-in goal control plane gives durable goals, event-sourced state, and verification gates to runs of 24+ hours — start a research task at night, review the results from your phone in the morning. Features Category Details Multi-Interface Terminal UI (TUI), desktop app (GUI), mobile apps (Android & iOS), CLI, IM bots — one agent, everywhere. The desktop backend also runs as a host on a server (future host: no window or WebView, works over SSH), and the phone↔desktop channel is end-to-end encrypted — a Noise handshake with ChaCha20-Poly1305 records, so the relay only carries ciphertext (host mode, remote) Self-Hosted Relay future relay init and future relay serve run the relay behind remote control yourself, with no FutureOS account in the path. The wire protocol, pairing and end-to-end encryption are unchanged, so the official relay and your own can run side by side (guide) Configurable Tool Safety read, write, edit, shell — approval rules and tiers (off / manual / sandbox, plus auto = sandbox with model review of the requests that would have asked, on a FutureOS sign-in); macOS Seatbelt, Linux system Bubblewrap, Windows restricted-token write protection. Availability and guarantees differ by platform (guide) Model Flexibility Bring your own key: the built-in catalog covers the major providers, and any OpenAI- or Anthropic-compatible endpoint can be added through models.json. Or sign in to the FutureOS provider (future auth login) and have the keys and model list provisioned for you. Both work side by side, and you choose which models the picker offers (setup) Goal Engineering Durable goals/todos/gates/monitors for long-horizon runs of 24+ hours — deterministic should-run kernel, event-sourced state, hard checks (evidence floor / acceptance contracts / verify gates), lease liveness, multi-agent (guide) Scheduled Tasks Recurring work without you in the loop: a reusable prompt + working directory + model/thinking level, run at full permission on a schedule (interval / daily / weekly / monthly / yearly, or a single date) — and on demand whenever you want. Every run is an ordinary conversation plus a run ledger with status, result summary and prompt-version history (guide) Powerful Built-in Skills 15+ skills out of the box for everyday agent work — image read & generation, PDF/Word parsing, web search, browser control, slides, software install, and the /future-goal long-run goal orchestrator (builtin) Forkable Sessions Branch any conversation like a repo — fork, clone, and tree navigation over JSONL session history Rust Core Agent, IM channel bridge, goal control plane, CLI, and TUI are all written in Rust — high performance with memory safety Quick Start Install One line, no source build required: macOS / Linux — one script, auto-detects the platform: macOS gets the official signed app (arm64 / Intel); Linux gets the .deb on Debian/Ubuntu (desktop app + unified future CLI) or the portable tarball everywhere else: curl -fsSL https://dl.future-os.cn/install.sh | bash Windows (PowerShell) — runs the signed installer silently: iex (irm https://dl.future-os.cn/install.ps1) Step-by-step installation for every platform (desktop app, toolchains, GUI packaging) is in the Build & Install guide. The installers finish by running future init. Both scripts check the downloaded artifact against the release manifest's SHA-256 before installing it, including when FUTUREOS_VERSION pins a release (that release's own latest.json is read). An artifact whose checksum cannot be established — no manifest, no sha256 for the platform, or a host without shasum/sha256sum — is refused instead of installed unverified. Only FUTUREOS_SKIP_VERIFY=1 ($env:FUTUREOS_SKIP_VERIFY = '1' on Windows) turns that check off, and it means installing whatever the mirror served you. Configure a model The agent needs at least one model provider before it can answer. The interactive setup supports both FutureOS and custom providers: future config FutureOS hosted models — device-flow sign-in provisions keys and a model list automatically: future auth login This works whether or not the agent is running: with the agent up the key is applied live; without it the key is written to ~/.future/agent/auth.json and picked up when the agent starts. Prefer your own keys? (BYOK & custom providers) Use a known provider. Put your API key in ~/.future/agent/auth.json, keyed by provider name. See the built-in model catalog — most providers have built-in base URLs and auto-discover their models: { "openai": { "type": "api_key", "key": "sk-..." } } For providers with user-specific base URLs (e.g. Azure's YOUR_RESOURCE), add a baseUrl field in auth.json: { "azure": { "type": "api_key", "key": "sk-...", "baseUrl": "https://my-resource.openai.azure.com/openai/v1" } } Custom provider. For providers not in the built-in catalog, specify everything in ~/.future/agent/models.json: { "providers": { "my-provider": { "apiKey": "sk-...", "baseUrl": "https://my-api.example.com/v1", "models": [ { "id": "my-model", "name": "My Model", "contextWindow": 128000, "maxTokens": 16384 } ] } } } Many self-hosted endpoints accept only some thinking levels (often just low/